Genpact and Google Cloud

Genpact Finance One – Revenue and PnL agents

Finance One is a suite of AI agents built on Google ADK and the A2A protocol with sub‑agents, enabling finance users to gain actionable insights from revenue and profit & loss data through natural language conversations in Google Gemini Enterprise.
 

Intelligent Routing Agent
 

This agent serves as the entry point, using a meta‑cognitive approach to analyze keywords and semantic intent in real time and route queries to the correct domain‑specific agent.

It ensures accurate and explainable routing while allowing users to analyze multiple financial domains without needing to know underlying data sources.
 

•  Classifies queries into Revenue or Profit & Loss domains
•  Dynamically builds and weights keyword dictionaries per query
•  Streams routing logic to the UI for transparency
•  Delegates full query context to the appropriate sub‑agent

Revenue Insights Agent
 

This agent converts natural language revenue questions into BigQuery Standard SQL, executes the query, and delivers business‑ready insights with recommendations and visualizations through a coordinated sub‑agent pipeline.
 

The agent supports analysis across regions, products, channels, customers, and profit centers, covering metrics like gross revenue, gross‑to‑net, trade promotions, BME spend, cash discounts, and contribution margins.

•  SQL Converter Agent: Generates optimized, read‑only BigQuery SQL using schema‑aware prompting, column mappings, and few‑shot examples.
•  Sequential Agent: Orchestrates schema retrieval, SQL generation, validation, execution, and result collection with error handling.
•  Query Answering Agent: Transforms results into clear summaries, formatted metrics, recommendations, and visuals such as tables, charts, and anomaly flags.

Profit & Loss Insights Agent
 

The P&L Insights Agent follows the same multi‑agent pattern to answer natural language questions about profitability, budgets, and costs using BigQuery‑backed analysis.
 

The agent enables analysis across GL accounts, cost centers, suppliers, plants, peg lines, and profit centers, highlighting actual vs. budget vs. forecast variances, allocation anomalies, and cost concentration risks.
 

•  SQL Converter Agent: Produces schema‑aware SQL targeting P&L segmentation and profit & loss tables.
•  Sequential Agent: Executes validated SQL through an ordered, fault‑tolerant pipeline.
•  Query Answering Agent: Generates summaries, variance analysis, cost optimization recommendations, and automated visualizations.